Sabermetrics Debunk Traditional Baseball Strategy
What if you managed your own real life baseball team? Draft your own players, decide the batting order, and make play by play decisions during games. Would you have your players stealing bases? Would you spend big bucks on a strong closing pitcher? Bill James would not advise you to do any of these things. In fact he has shown that a number of common decisions made in baseball do not make statistical sense. Why should you listen to him, you ask?
Bill James has been studying baseball statistics for years and started publishing his findings in 1977 [1]. He coined the term sabermetrics, which can be defined as the mathematical and statistical analysis of baseball. Baseball, like many sports, is driven by statistics. traditional thinking often puts an emphasis on stats that everyone believes to be important, such as batting average and earned runs average. But sabermetrics takes statistics one step further.
Bill James devises formulas that evaluate a player’s pure ability to add or prevent scores in a game. The ‘On Base Percentage’ has a very high importance and when coupled with other stats, a player’s value to his team can be represented as a single number.
Sabermetrics can also shift a strategy’s focus away from popular statistics and towards more vital, fundamental ones. The ‘out’ takes on colossal value in sabermetrics, and an offensive manager’s strategy can be modeled around keeping them from happening [2].
Sports journals report on the trends they see, fans debate amongst one another, and everyone forms opinions, discounting information they don’t want to see. Typical Major League Baseball managers make about 11,000 decisions during the course of a season [2], many of them made on gut feelings. They might see statistics, over-value some and under-value others. By applying sabermetrics to baseball statistics, several common decisions made during the course of the game are revealed to be poor ones.
Sacrifice Hits
Batters sometimes sacrifice themselves in order to move a runner already on base (I am referring to bunts rather than unintentional sacrifice flies). However, the out lost is much more valuable than the base gained. “Baseball’s Secret Formula” gives us the following scenario when applying BillJames’s sabermetrics: a runner on first with no outs gives the batting team an average of .95 runs per inning. When the batter sacrifices his at-bat, it gives the team a runner on second and one out, but the team’s average runs per inning drops to .73 for the inning. A runner on second with two outs further plummets the team’s scoring average to .25 for the inning. However, if the batter hits a single instead of sacrificing, the team’s runs per inning soars to at least 1.57 runs with runners on first and second with no outs [2].
Stealing Bases
Stealing bases is another common scenario in baseball games that does not make sense in the long run. Major league baseball players have roughly a 70% percent success rate when attempting to steal a base. Yet when they are tagged out, the team’s chances of scoring a run in that inning plummets. Whatever short-term positive effect a successful steal might have is greatly outweighed by the taking this unnecessary risk and negative effect of a failed steal attempt [2]. Additionally, there is no correlation between stolen bases and wins.
Walking A ‘Star’ Batter
Afraid Barry Bonds will hit one out of the park? Why not intentionally walk him? The basic principle here is to avoid allowing specific batters to hit, fearing a scoring play. However, when a pitcher puts a man on base, he effectively increases batting team’s chances of scoring in that inning. This is true regardless of who is on base and who else bats that inning.
Each each of these examples, sabermetrics values runners on base with as few outs as possible (opposite if you’re pitching). Some traditional strateiges try to earn short term gain, but sabermatrics will never intentionally put a man on base or unnecessairly risk an out. Over the course of a season, it doesn’t pay.
Other Notable Findings
While sabermetrics has been around since the late 1970′s, there does not seem to be much written about it outside of the immediate community. This site extracted some more notable findings from a 1988 book by James. Another individual usedsabermetrics to conclude that ‘clutch hitters’ don’t exist, and wouldn’t be significant anyway. In the Science Channel’s “Baseball’s Secret Formula” [2], Bill James reveals several other conclusions:
- Going into the ninth inning with a two run lead, a team has a 95% change of winning no matter what happens.
- Strong closing pitchers are most effective in the seventh inning when a team is tied or down a run, rather than in the ninth when they’re up by one or two.
- Making subtle changes in a batting lineup does not radically change the amount of runs a team produces over time.
So If The Stats Tell The Truth, Why Aren’t MLB Teams Leveraging Them?
Traditional baseball coaches may stick with inaccurate methodologies involving intuition, but younger, modern coaches are starting to catch on. Teams in other sports have been doing it for years. Some teams have been rather successful (see RedSox in 2004 http://www.redsoxstats.com/).
Citations
- (2005, June 28) . Retrieved November 2, 2007 “Bill James, Beyond Baseball” PBS Think Tank with Ben Wattenberg
- (2008 ,February 14) Retrieved February 14, 2008 “Baseball’s Secret Formula”. The Science Channel.
You should follow us on twitter.
As to the sacrifice bunt, one of the primary reasons it’s employed is to stay out of a double play. Weak hitters (notably pitchers) have a propensity for doubling up a runner at first. While it may be true that “a runner on first with no outs gives the batting team an average of .95 runs per inning,” I’m guessing a team with two outs and no one on base has a considerably lower chance to score runs that inning. Most pitchers bat well under .200 in a season (most actually hit around .100 – .130), and 90 percent of their outs are either Ks or ground outs. A strikeout with a runner on first yields no benefit. A ground out with a runner on first has a high probability of turning into two outs. Under those circumstances, isn’t it better to bunt?
Zac, you raise a good point. Logically, it makes sense to avoid allowing pitchers to cause the team damage when batting. These stats were taken across several years of games, including all teams and players. If we narrowed this sample down to, say, just the seventh, eighth and ninth batters, those numbers might be significantly different.
This aside, you may be introducing somewhat of a new topic in your comment above. The purpose of a sacrifice bunt that I described is to advance a runner already on base. I think you’re talking about sacrificing to defend this runner (protecting him from getting out).
When a batter sacrifices himself, the manager is conceding that he can’t defend him from an out. So why not advance a runner in the process, right? But sabermetrics might tell you to let the batter swing away anyway. He may strike out or hit into a double play, but when he does get a hit, the positives outweigh the negatives.
But again, this metric refers to long-term results involving all players. Thanks for the input! I’d be interested to hear more…
Co-worker of mine offered this via instant message:
Very interesting view on baseball logic
Pingback: Noobie Game Guide? - OOTP Developments Forums
I love the application of statistical analysis to the game of baseball, and the resulting dispelling of some baseball ‘myths’. However we have to remember that Sabermetrics in a Macro examination of the game, but while playing the game we are involved in a Micro analysis for the strategy applied.
So let’s take the sacrifice bunt again.
I am down one run 2-3 in the bottom of the last inning.
I have Tommy on first no outs.
At bat is Johnny, batting .200 on the year and .125 against the current pitcher.
The on deck batter and in hole batter both bat over .300.
Tommy steals bases at a .800 clip. The catcher has a weak arm.
The pitcher throws ball one.
The pitcher throws ball two. 2-0 is the count. Johnny still only bats .125 on this pitcher with a batter favorable count.
THIS IS A MICRO SITUATION THAT FACES A COACH. Here is my analysis as the COACH.
At 2-0 the pitcher has to throw a strike or attempt to throw a strike, so the pitch-out is out of the question. In this situation i estimate with the weak arm of the catcher, my guy on first – Tommy would be successful (barring falling down) almost 100% of the time.
So i send Tommy on the 2-0 pitch, he steals successfully and the pitch is a strike (batter fake bunts to hold the catcher in the box).
Now 2-1 count, i sacrifice bunt Johnny moving Tommy to third base.
I now have one out a man on third and two consecutive batters over .300 coming to the plate.
Didn’t i just use my data to give my team the best chance of tying or winning the game.
I think we error in confusing MACRO with MICRO analysis of data. MACRO analysis give us the general information for the game, but the game is PLAYED IN THE MICRO. As a Coach i shouldn’t be afraid of a statistical analysis of the game, but i should be smart enough to use the MICRO analysis on the field during play.
Coach Wright
My comment is similar to Coach Wright’s comment. Sabermetric based strategies typically compare the average number of runs scored in an inning based on one decision, to the average number of runs scored using another. While I agree that most of the time this is the correct way to evaluate a strategy, I would argue that there are some situations where it is not. for example:
From the sacrifice bunt example above, having a runner on first base with no outs is worth an average of .95 runs. Having a runner on second base with one out is worth an average of .75 runs. However, this does not mean that you should never sacrifice bunt.
If I am in a tie game in the bottom of the ninth with a man on first and nobody out, the appropriate question is not “Which decision will give me on average the most runs”. The appropriate question is “Which decision gives me the highest probability of scoring at least one run?”
If you look not just at the average number of runs scored, but the distribution of runs scored, you will see that the number .95 is influenced by the fact that you are much more likely to have a “big inning” with no outs and a guy on first than you are if you have one out and a guy on second. The high potential for big innings makes the average value for the non-bunting scenario higher than the non-bunting scenario. However, scoring more than one run is irrelevant in the bottom of the ninth in a tie game. Although I haven’t seen the data, I would be willing to bet that you are more likely to score 1 or more runs by sacrificing, even though, on average you will score more runs by letting the hitter hit.
Coach Briggs